1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
|
<div class="subsection-level-extent" id="Nvidia-PTX-Options"> <div class="nav-panel"> <p> Next: <a href="openrisc-options" accesskey="n" rel="next">OpenRISC Options</a>, Previous: <a href="nios-ii-options" accesskey="p" rel="prev">Nios II Options</a>, Up: <a href="submodel-options" accesskey="u" rel="up">Machine-Dependent Options</a> [<a href="index#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="indices" title="Index" rel="index">Index</a>]</p> </div> <h1 class="subsection" id="Nvidia-PTX-Options-1"><span>3.19.35 Nvidia PTX Options<a class="copiable-link" href="#Nvidia-PTX-Options-1"> ¶</a></span></h1> <p>These options are defined for Nvidia PTX: </p> <dl class="table"> <dt>
<span><code class="code">-m64</code><a class="copiable-link" href="#index-m64"> ¶</a></span>
</dt> <dd>
<p>Ignored, but preserved for backward compatibility. Only 64-bit ABI is supported. </p> </dd> <dt>
<span><code class="code">-march=<var class="var">architecture-string</var></code><a class="copiable-link" href="#index-march-12"> ¶</a></span>
</dt> <dd>
<p>Generate code for the specified PTX ISA target architecture (e.g. ‘<samp class="samp">sm_35</samp>’). Valid architecture strings are ‘<samp class="samp">sm_30</samp>’, ‘<samp class="samp">sm_35</samp>’, ‘<samp class="samp">sm_53</samp>’, ‘<samp class="samp">sm_70</samp>’, ‘<samp class="samp">sm_75</samp>’ and ‘<samp class="samp">sm_80</samp>’. The default depends on how the compiler has been configured, see <samp class="option">--with-arch</samp>. </p> <p>This option sets the value of the preprocessor macro <code class="code">__PTX_SM__</code>; for instance, for ‘<samp class="samp">sm_35</samp>’, it has the value ‘<samp class="samp">350</samp>’. </p> </dd> <dt>
<span><code class="code">-misa=<var class="var">architecture-string</var></code><a class="copiable-link" href="#index-misa"> ¶</a></span>
</dt> <dd>
<p>Alias of <samp class="option">-march=</samp>. </p> </dd> <dt>
<span><code class="code">-march-map=<var class="var">architecture-string</var></code><a class="copiable-link" href="#index-march-13"> ¶</a></span>
</dt> <dd>
<p>Select the closest available <samp class="option">-march=</samp> value that is not more capable. For instance, for <samp class="option">-march-map=sm_50</samp> select <samp class="option">-march=sm_35</samp>, and for <samp class="option">-march-map=sm_53</samp> select <samp class="option">-march=sm_53</samp>. </p> </dd> <dt>
<span><code class="code">-mptx=<var class="var">version-string</var></code><a class="copiable-link" href="#index-mptx"> ¶</a></span>
</dt> <dd>
<p>Generate code for the specified PTX ISA version (e.g. ‘<samp class="samp">7.0</samp>’). Valid version strings include ‘<samp class="samp">3.1</samp>’, ‘<samp class="samp">6.0</samp>’, ‘<samp class="samp">6.3</samp>’, and ‘<samp class="samp">7.0</samp>’. The default PTX ISA version is 6.0, unless a higher version is required for specified PTX ISA target architecture via option <samp class="option">-march=</samp>. </p> <p>This option sets the values of the preprocessor macros <code class="code">__PTX_ISA_VERSION_MAJOR__</code> and <code class="code">__PTX_ISA_VERSION_MINOR__</code>; for instance, for ‘<samp class="samp">3.1</samp>’ the macros have the values ‘<samp class="samp">3</samp>’ and ‘<samp class="samp">1</samp>’, respectively. </p> </dd> <dt>
<span><code class="code">-mmainkernel</code><a class="copiable-link" href="#index-mmainkernel"> ¶</a></span>
</dt> <dd>
<p>Link in code for a __main kernel. This is for stand-alone instead of offloading execution. </p> </dd> <dt>
<span><code class="code">-moptimize</code><a class="copiable-link" href="#index-moptimize"> ¶</a></span>
</dt> <dd>
<p>Apply partitioned execution optimizations. This is the default when any level of optimization is selected. </p> </dd> <dt>
<span><code class="code">-msoft-stack</code><a class="copiable-link" href="#index-msoft-stack"> ¶</a></span>
</dt> <dd>
<p>Generate code that does not use <code class="code">.local</code> memory directly for stack storage. Instead, a per-warp stack pointer is maintained explicitly. This enables variable-length stack allocation (with variable-length arrays or <code class="code">alloca</code>), and when global memory is used for underlying storage, makes it possible to access automatic variables from other threads, or with atomic instructions. This code generation variant is used for OpenMP offloading, but the option is exposed on its own for the purpose of testing the compiler; to generate code suitable for linking into programs using OpenMP offloading, use option <samp class="option">-mgomp</samp>. </p> </dd> <dt>
<span><code class="code">-muniform-simt</code><a class="copiable-link" href="#index-muniform-simt"> ¶</a></span>
</dt> <dd>
<p>Switch to code generation variant that allows to execute all threads in each warp, while maintaining memory state and side effects as if only one thread in each warp was active outside of OpenMP SIMD regions. All atomic operations and calls to runtime (malloc, free, vprintf) are conditionally executed (iff current lane index equals the master lane index), and the register being assigned is copied via a shuffle instruction from the master lane. Outside of SIMD regions lane 0 is the master; inside, each thread sees itself as the master. Shared memory array <code class="code">int __nvptx_uni[]</code> stores all-zeros or all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD regions). Each thread can bitwise-and the bitmask at position <code class="code">tid.y</code> with current lane index to compute the master lane index. </p> </dd> <dt>
<span><code class="code">-mgomp</code><a class="copiable-link" href="#index-mgomp"> ¶</a></span>
</dt> <dd>
<p>Generate code for use in OpenMP offloading: enables <samp class="option">-msoft-stack</samp> and <samp class="option">-muniform-simt</samp> options, and selects corresponding multilib variant. </p> </dd> </dl> </div> <div class="nav-panel"> <p> Next: <a href="openrisc-options">OpenRISC Options</a>, Previous: <a href="nios-ii-options">Nios II Options</a>, Up: <a href="submodel-options">Machine-Dependent Options</a> [<a href="index#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="indices" title="Index" rel="index">Index</a>]</p> </div><div class="_attribution">
<p class="_attribution-p">
© Free Software Foundation<br>Licensed under the GNU Free Documentation License, Version 1.3.<br>
<a href="https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Nvidia-PTX-Options.html" class="_attribution-link">https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Nvidia-PTX-Options.html</a>
</p>
</div>
|
